Understanding admixture analysis

Emil O. W. Kirkegaard, John Fuerst

Large group differences in intelligence exist, but why?

Group differences are to large extent racial in nature

Racial group differences are very stable over time

Big debate over causes! - expert survey data

Debate over whether mixed populations follow the pattern expected by genetics

• Simple genetic models (additive) predict linear relationship
• But some old studies seem to not find this pattern
• Witty and Jenkins (1936) studied gifted black kids, and supposedly found no elevated European ancestry
• Eyferth (1961) based on kids from post-WW2 German mothers and US soldiers -- both blacks and whites
• Scarr et al (1977) used blood groups to estimate African ancestry in US blacks, correlated with IQ but found nothing much

But evidence quality is poor

• Single studies are suspect -- studies not replicated ever
• Often small sample sizes, e.g. gifted blacks kids n = 63
• Ancestry estimates dubious
• Family history is often wrong
• Blood group analysis very primitive by modern standards
• Very likely major publication bias and probably QRPs
• Bias direction follows researcher political biases
• Lots of more old data exist, but usually ignored
• No one bothered to update the reviews of the data since Jensen (1973)

Genomics is upon us

• Easy to measure genomic ancestry using DNA
• Fairly cheap, about 100 USD/person from DTC companies
• Large-scale commercial price is about 25 USD/person

But, to really understand admixture analysis, need to go back to basics

• 10 loci, each with .50 frequency
• No group difference in frequencies
• Many loci means distribution becomes normal/Guassian, cf. central limit theorem.

Add some noise and we get a normal distribution

Add small difference in frequency of each locus

• 10 loci + noise
• Group frequencies of .55 and .45 at each locus
• Group means 4.5 and 5.5

Now mate people at random

Large datasets with genomics exist

• Some of these have IQ, skin tone, self-reported discrimination
• Some even have (almost) all of these, e.g. UKBB (UK), Pelotas (Brazil), many US ones
• However, data not public, locked away
• Strict access rules -- only professors can apply for them
• And I’m not a professor…
• But if try anyway...

“I asked Prof. X about this project, and even though he does recognise its relevance, I am afraid that he declined to provide the data. The reason is that country C is facing a very delicate political situation at the moment, and race/ethnicity/ancestry is one of the topics at the core of debates and etc. Everyone in the country seems to be a little bit cautious when it comes to looking at ethnicity, especially regarding this such as intelligence or violence.

I am truly sorry that we will not be able to help at this opportunity, but I do wish the best of luck with your research.”

However, we did find some data...

• Noble et al (2015)
• PING dataset, n = 1500 US kids, about 50% non-white, 1400 with genomic ancestry data
• IQ data, brain data, parental SES -- no skin tone, no discrimination self-report
• Pelotas 1982 dataset via de Franca et al (2017)
• N = 3000ish, Brazil birth cohort 1982, highly admixed population
• IQ, education, income, skin tone, discrimination self-report
• But we don’t have the IQs or income or discrimination self-report!
• Both datasets found via tips from online HBDers
• Might be more to find, keep looking!

PING basic genomic ancestry

PING main model

PING, control for parental SES

Visualizing the results - 2d

n = 140 African Americans with 2-way ancestry

Visualizing the results - 3d

Pelotas - European ancestry and education

• Public dataset only has ancestry quintiles not precise values.

Pelotas regression

Pelotas regression, with skin tone

More studies on the way

• At least one other researcher is also working on an ancestry study
• Problems getting published
• Consistent results for IQ for US data, but small sample, about 200
• The need for measures of ‘non-genetic’ explanations
• Skin tone/discrimination
• But even then interpretation is questionable because reporting discrimination is likely influenced by media/culture and personality/cognitive traits
• More definitive genomic methods exist, e.g. admixture mapping

​

References

• Kirkegaard, E., Fuerst, J., & Meisenberg, G. (2018, April 26). Genomic ancestry, cognitive ability, and socioeconomic outcomes. http://doi.org/10.17605/OSF.IO/4AN93
• Preprint, beware 70 pages!
• Simulation output http://rpubs.com/EmilOWK/how_to_genetic_group_differences
• de Franca et al. (2017). Genomic ancestry and education level independently influence abdominal fat distributions in a Brazilian admixed population. PLOS1. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5459508/
• Noble et al. (2015). Family income, parental education and brain structure in children and adolescents. Nature Neuroscience. doi:10.1038/nn.3983
• Eyferth study https://en.wikipedia.org/wiki/Eyferth_study
• Original German: http://emilkirkegaard.dk/en/wp-content/uploads/Eyferth-1961.pdf
• Scarr et al. (1977). Absence of a relationship between degree of white ancestry and intellectual skills within a black population. 10.1007/BF00273154
• Witty and Jenkins (1936). Intra-race testing and negro intelligence. The Journal of Psychology: Interdisciplinary and Applied, 1, 179-192.